This document discusses a GNU Radio-based full duplex radio system. It begins with an overview of full duplex communications and the challenges of self-interference. It then describes how digital cancellation works, including modeling the nonlinear distortions in the transmission path and using algorithms like least mean squares to solve the cancellation problem. The document outlines both block-based and adaptive algorithms that can be used for digital self-interference cancellation.
This document discusses a GNU Radio-based full duplex radio system. It begins with an overview of full duplex communications and the challenges of self-interference. It then describes how digital cancellation works, including modeling the nonlinear distortions in the transmission path and using algorithms like least mean squares to solve the cancellation problem. The document outlines both block-based and adaptive algorithms that can be used for digital self-interference cancellation.
This document discusses a GNU Radio-based full duplex radio system. It begins with an overview of full duplex communications and the challenges of self-interference. It then describes how digital cancellation works, including modeling the nonlinear distortions in the transmission path and using algorithms like least mean squares to solve the cancellation problem. The document outlines both block-based and adaptive algorithms that can be used for digital self-interference cancellation.
This document discusses a GNU Radio-based full duplex radio system. It begins with an overview of full duplex communications and the challenges of self-interference. It then describes how digital cancellation works, including modeling the nonlinear distortions in the transmission path and using algorithms like least mean squares to solve the cancellation problem. The document outlines both block-based and adaptive algorithms that can be used for digital self-interference cancellation.
– What is Full Duplex? – Implementation – The Problem – Cancellation Performance – The Solution – Throughput Results
• Theory: Digital Cancellation • Using USRP Products
– Black Box Model – Timing Synchronization – Modeling Nonlinearity – Built-in vs. External Mixers – Least-squares Problem – Algorithms for Solving • Conclusions
2 What is Full-Duplex Communications? … it’s not the ‘usual definition’
• Transmit and receive on the same antenna on the same frequency at the same time – Potentially double (or more) throughput in the same spectrum – Considered impossible for typical communications links – Multiple classical textbooks explicitly state it can’t be done
• Can replace both time- and frequency-multiplexed links
Traditional Time-Division Multiplexing Traditional Frequency-Division Multiplexing
Guard Band Uplink Downlink Uplink Downlink Uplink Downlink
time frequency Full-Duplex Link Full-Duplex Link
Uplink Uplink
Downlink Downlink
time frequency
3 The Problem
• The problem is self-interference
– Transmit power swamps the receive power, making it very difficult to detect the desired signal – Power difference 10log10(PTx/PRx) depends on the distance between Tx and Rx
TX D/A PA Self- interference
Sources of self-interference • Circulator leakage RX A/D LNA • Antenna reflection • EM coupling on cables • Reflections off objects
4 The Solution Very active area of research in the past several years
• The transmitted signal is known at the receiver
– To some degree, need to account for delay and distortion(s) • Subtract the transmitted signal from the received signal to remove self-interference • The devil is in the implementation – Transmitted signal power may be > 100 dB above received signal – Need very high linearity, very accurate matching and model of distortions • Two major approaches: 1. Cancel at baseband in the digital domain 2. Cancel at RF in the analog domain
TX D/A PA Self- interference Digital RF Cancellation Cancellaton
RX A/D LNA
All reported full-duplex systems use both analog and digital cancellation 5 Digital Cancellation
• Find a model for the black box to minimize r.
– I.e. find ℱ to minimize |𝑟| = 𝑏 − ℱ 𝑎 • The model must account for: – Gain – Phase shift – Time delay – Nonlinear distortion – Multipath
6 Modeling the Black Box Linear Model
• We model the black box as a non-causal FIR filter.
• Let: 𝑎𝑖 = transmitted signal, sample 𝑖 𝑏𝑖 = received signal, sample 𝑖 𝑥𝑘 = FIR filter coefficient 𝑘 • Our goal is to find the filter coefficients that produce outputs (ri) most similar to the actual received signal. • Example for a 5-tap filter, for the 2nd received sample: 𝑏2 = 𝑥0 𝑎0 + 𝑥1 𝑎1 + 𝑥2 𝑎2 + 𝑥3 𝑎3 + 𝑥4 𝑎4 • Or, in vector form: 𝑥0 𝑥1 𝑎0 𝑎1 𝑎2 𝑎3 𝑎4 𝑥2 = 𝑏2 𝑥3 𝑥4
7 Modeling the Black Box Nonlinear Distortion
• We model the black box as a summation of FIR filters, each
operating on a different (odd) power of the transmitted signal. • Let: 𝑎𝑖 = transmitted signal, sample 𝑖 𝑥10 𝑥11 𝑏𝑖 = received signal, sample 𝑖 𝑥12 𝑥𝑗𝑘 = FIR filter coefficient 𝑘 for power 𝑗 𝑥13 • Example for 3 powers (1st, 3rd,and 5th), each with 5 taps: 𝑥14 𝑥30 𝑥31 𝑎0 𝑎1 𝑎2 𝑎3 𝑎4 𝑎03 𝑎13 𝑎23 𝑎33 𝑎43 𝑎05 𝑎15 𝑎25 𝑎35 𝑎45 𝑥32 = 𝑏2 𝑥33 𝑥34 𝑥50 𝑥51 𝑥52 𝑥53 𝑥54 8 Modeling the Black Box Matrix Representation 𝑥10 • If we wish to work with multiple samples of the received signal at a time, 𝑥11 we can express the model in matrix form as follows: 𝑥12 𝑥13 𝑎0 𝑎1 𝑎2 𝑎3 𝑎4 𝑎03 𝑎13 𝑎23 𝑎33 𝑎43 𝑎05 𝑎15 𝑎25 𝑎35 𝑎45 𝑥14 𝑏2 𝑎1 𝑎2 𝑎3 𝑎4 𝑎5 𝑎13 𝑎23 𝑎33 𝑎43 𝑎53 𝑎15 𝑎25 𝑎35 𝑎45 𝑎55 𝑥30 𝑏3 𝑎2 𝑎3 𝑎4 𝑎5 𝑎6 𝑎23 𝑎33 𝑎43 𝑎53 𝑎63 𝑎25 𝑎35 𝑎45 𝑎55 𝑎65 𝑥31 𝑏4 𝑎3 𝑎4 𝑎5 𝑎6 𝑎7 𝑎33 𝑎43 𝑎53 𝑎63 𝑎73 𝑎35 𝑎45 𝑎55 𝑎65 𝑎75 𝑥32 = 𝑏5 𝑥33 𝑏6 𝑎4 𝑎5 𝑎6 𝑎7 𝑎8 𝑎43 𝑎53 𝑎63 𝑎73 𝑎83 𝑎45 𝑎55 𝑎65 𝑎75 𝑎85 𝑥34 𝑏7 𝑎5 𝑎6 𝑎7 𝑎8 𝑎9 𝑎53 𝑎63 𝑎73 𝑎83 𝑎93 𝑎55 𝑎65 𝑎75 𝑎85 𝑎95 𝑥50 ⋮ ⋮ 𝑥51 𝑥52 𝑥53 𝑥54 𝐴 𝑥 = 𝑏 We are looking for the value of 𝑥 that minimizes 𝐴𝑥 − 𝑏 9 Algorithms For Solving 𝐴𝑥 = 𝑏
Block-based algorithms operate on the Adaptive algorithms operate on one row of
entire matrix 𝐴, effectively computing 𝐴 at a time, adjusting the value of 𝑥 each 𝑥 = 𝐴+ 𝑏. iteration.
• Singular value decomposition (SVD) • Least mean squares (LMS)
• QR decomposition • Recursive least squares (RLS) • Cholesky decomposition • Conjugate gradient method
10 Algorithms For Solving 𝐴𝑥 = 𝑏
Block-based algorithms operate on the Adaptive algorithms operate on one row of
entire matrix 𝐴, effectively computing 𝐴 at a time, adjusting the value of 𝑥 each 𝑥 = 𝐴+ 𝑏. iteration.
• Singular value decomposition (SVD) • Least mean squares (LMS)
• QR decomposition • Recursive least squares (RLS) • Cholesky decomposition • Conjugate gradient method
We implemented a GNU Radio block for each algorithm in green.
11 Note
• These GNU Radio blocks are not just for full duplex!
• They can be used to solve any problem where it is desirable to cancel a known signal that may have undergone linear and/or nonlinear distortion.
12 Performance Optimization
• For high throughput performance, we leverage the Intel libraries:
– IPP: Intel Performance Primitives:
• Contains signal processing routines • Optimized using Streaming SIMD Extensions
– MKL: Math Kernel Library
• Contains optimized functions for vector and matrix math • API is compatible with BLAS and LAPACK functions
• We further increase throughput by dividing computationally-intensive work
between multiple threads.
13 Implementation (Pseudocode) SVD Block
int svd_canceller_cc_impl::general_work( int noutput_items, gr_vector_int &ninput_items, gr_vector_const_void_star &input_items, gr_vector_void_star &output_items) { gr_complex* ref = input_signals[0]; // ref = transmitted signal gr_complex* b = input_signals[1]; // b = received signal gr_complex* out = output_signals[0];
construct_A_matrix(ref, A); cgelsd(A, b, x, ...); // solve Ax = b for x (using SVD) cgemv(A, x, b, out, ...); // residual = b – Ax
consume_each(block_size); return block_size; }
14 Implementation (Pseudocode) QR Decomposition Block
int qr_canceller_cc_impl::general_work( int noutput_items, gr_vector_int &ninput_items, gr_vector_const_void_star &input_items, gr_vector_void_star &output_items) { gr_complex* ref = input_signals[0]; // ref = transmitted signal gr_complex* b = input_signals[1]; // b = received signal gr_complex* out = output_signals[0];
construct_A_matrix(ref, A); cgeqrf(A, temp, ...); // perform QR factorization on A cungqr(temp, Q, ...); // compute Q explicitly cgemv(Q, b, Qb, ...); // compute 𝑄 𝑇 𝑏 cgemv(Q, Qb, b, out, ...); // residual = 𝑏 − 𝑄 𝑄𝑇 𝑏
consume_each(block_size); return block_size; }
15 Implementation (Block Diagram) LMS Block
• We optimize this block for throughput by:
– Using multiple threads, each of which processes a chunk of the input – Calculating updates to the coefficients 𝑥 every N samples (instead of every sample) – Averaging the threads’ values of 𝑥 when they synchronize
• For this flowgraph to work, the USRP sink and source blocks must start streaming at the same time. • These blocks provide functions for timing synchronization: – set_time_now() – set_time_next_pps() – set_start_time() • We can edit the GRC-generated Python code to call these functions ... – But if we make a change to the flowgraph and regenerate, we have to do it again – There must be a better way ...
20 Function Caller Block
• Wouldn’t it be nice if there was a GRC block that allowed us to embed arbitrary function calls in the generated code, that execute before the flowgraph starts?
• We created one: Generated Python code: val = self.usrp_source.set_time_now(0) ... val = self.usrp_sink.set_time_now(0) ... val = self.usrp_source.set_start_time(1) ... val = self.usrp_sink.set_start_time(1)
• Largely based on the built-in Function Probe block
21 Using USRP Products Timing Synchronization
• Even after synchronizing the USRP source and sink, there is still a timing offset: – Deterministic and repeatable – Appears to be sample-rate dependent – On the order of 10-50 samples
• This can be remedied with a skip head block:
22 Using USRP Products Built-in vs. External Mixers
• Most USRP daughterboards contain LO • Some daughterboards contain no LO or mixer but
generators and mixers to convert between operate at baseband/IF. baseband/IF and RF. • e.g. BasicTX, BasicRX • e.g. SBX, UBX
